Skip to content
Audio

Prakriti v1

Emotional TTS for Hindi, Hinglish, and English

2 min readupdated

This new 82M-parameter text-to-speech (TTS) model builds on the popular Kokoro-82M architecture and is trained for Hindi, mixed Hinglish, and Indian English. It supports 10 distinct voices and emotional style controls (joy, sadness, anger, fear, surprise, and more). Retaining high generation speeds—around 36x real-time on a GPU and up to 2.6x across four CPU threads—it is a strong fit for real-time voice assistants.

Key features

  • 10 unique voices: 2 professionally recorded in a studio and 8 distilled from a larger teacher model.
  • Customizable emotions and intensity levels for every voice.
  • Advanced Hindi text engine: accurately handles numbers, currencies, and time expressions, alongside schwa deletion and a pronunciation dictionary.
  • Sentence-level streaming support for instant playback on both CPU and GPU.

What you can do

  • Generate expressive speech and dialogue in Hindi and English for bots or apps with tailored emotional tones.
  • Build a fast local speech synthesis service thanks to its high throughput on standard hardware.
  • Use streaming output to power responsive, low-latency voice assistants.

Example

from prakriti.tts import Prakriti
import soundfile as sf

tts = Prakriti.from_pretrained("anand-k-maurya/prakriti-v1")
audio = tts.synthesize("नमस्ते! आपका order कल तक पहुँच जाएगा।", voice="Rasa_F", emotion="happy", intensity=0.8)
sf.write("out.wav", audio, 24000)
Read moreDetails
• Base: Kokoro-82M (fine-tuned using the StyleTTS 2 recipe)
• Dataset: 54.7 hours after filtering (including Rasa data and Maya 2 synthetic speech)
• Hardware: Requires a single RTX A6000 for training; runs on both GPU and CPU for inference
• License: Apache-2.0
Discuss Telegram 1 22

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.3 GB · ready files0.8 GB
  • FP8 / INT8recommendedweights 0.2 GB0.7 GB
  • 4-bitweights 0.1 GB0.6 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # anand-k-maurya/prakriti-v1
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download anand-k-maurya/prakriti-v1 --local-dir ~/models/prakriti-v1
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to prakriti-v1
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. StyleTTS2-LJSpeechfamily root
  2. Kokoro-82M
  3. prakriti-v1

New posts about versions of StyleTTS2-LJSpeech (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 3All on Hugging Face
Fine-tunes 12All on Hugging Face

About the model

anand-k-maurya/prakriti-v1
Author
anand-k-maurya
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

221

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Audio

Qwen3-ASR

Qwen3-ASR